Enjoy the full story in our Communications Earth & Environment paper: “Non-randomness of Japan megaquakes implied by stress recovery and accumulation.”

Our analysis combines several strands of information. First, we estimate spatially and temporally varying b-values along the plate boundary where the Pacific plate subducts beneath northeastern Japan, using earthquakes recorded by the Japan Meteorological Agency since 2000. Laboratory experiments and field observations suggest that low b-values tend to occur in regions of relatively high differential stress, whereas high b-values are associated with lower stress. In that sense, b can serve as a proxy for stress on the fault.

Second, we compare the b-value patterns with independent evidence on fault loading and release. This includes interplate coupling inferred from GNSS observations, patterns of seismic quiescence for moderate earthquakes, and the distribution of slow earthquakes such as tremors and very low frequency events. By analyzing all of these together, we aim to build a more coherent picture of where stress is accumulating and where it has been largely released.

At the beginning of the project we were not sure whether the data would really “speak” clearly. Immediately after a great earthquake, seismicity can be highly variable, and there is always a risk of over-interpreting short-term fluctuations. One turning point came when we extended our b-value timeseries all the way to June 2025 and saw that the high b-values in the Tohoku source region persisted , while the offshore Hokkaido region showed a long, steady trend toward lower b (Figure 1b,c). That contrast gave us confidence that we were seeing a robust, physically meaningful signal rather than noise.

The results tell two different stories for Tohoku and Hokkaido. In the source region of the 2011 M9 Tohoku earthquake, we find that b-values were low before the mainshock (Figure 1a,c), consistent with a high-stress state prior to failure. After the earthquake, b-values increase and remain high at least until June 2025 (Figure 1b,c), indicating a persistently low-stress state. This behaviour is inconsistent with a rapid recovery of stress to pre-2011 levels. It suggests that another M9-class event on the same segment is unlikely in the near term and that a long period of stress accumulation will be required before such an event can recur. However, this does not eliminate the possibility of M7–8-class earthquakes and tsunamis, which remain important components of regional hazard.

Offshore Hokkaido, the picture is quite different. East of the high-slip area of the 2003 M8 Tokachi earthquake, we identify a region where b-values are low and have decreased further since around 2008 (Figure 1b,c). This low-b region coincides with a zone of reduced rates of moderate earthquakes and strong plate coupling. It also overlaps with fault models of the 17th-century tsunami earthquake, whose magnitude is estimated to be around M8.8, and with a recognised seismic gap along the Japan trench. The combination of low b-values, seismic quiescence, and strong coupling suggests that this part of the margin is in a relatively high-stress state.

This interpretation is broadly consistent with long-term evaluations that estimate a 340–380 year recurrence interval for the 17th-century-type event and note that roughly 400 years have now passed. While our study cannot predict the timing of any individual earthquake, it does indicate that, in a relative sense, the offshore Hokkaido segment is much closer to failure than the Tohoku segment that ruptured in 2011. For coastal communities facing the Pacific in Hokkaido and northern Tohoku, this reinforces the importance of long-term tsunami preparedness, including evacuation planning and community awareness.

Behind the technical details, this work reflects many years of collaboration between seismologists and geodesists at the University of Shizuoka, JAMSTEC, and other institutions. As a seismologist who has long worked on statistical indicators such as the b-value, it has been both challenging and rewarding to integrate these indicators with geodetic coupling models and geological constraints from tsunami deposits. There were many discussions – sometimes late at night – about how to reconcile different datasets and how cautiously we should phrase our conclusions, given their potential societal implications.

Our hope is that this kind of integrative approach can help move the discussion of megaquake hazard from “anywhere, anytime” toward a more nuanced understanding of which segments are closer to rupture on the timescales that matter for societal planning. Ultimately, the goal is not to provide dramatic predictions, but to offer scientifically grounded guidance about where to prioritise long-term mitigation efforts. Small earthquakes are often overlooked, but when analysed carefully, they offer a powerful window into the evolving stress field of subduction zones – and thus into the future of the largest earthquakes that society fears most.